AI Race vs Human Flourishing: What US-China Talks Miss

Nate B. Jones (0:00)

I am shocked by how much of a barrier that sense of willingness is. Like, I talked to a lot of folks in the business, in enterprises, and they're like, I don't, like, my engineers care about Codex. That's fine for them. I'm not going to make everyone a coder in my company. I have other things to do with my business. When you look at the next 10 years or so, paint me, I guess maybe paint me two paths, right? Paint me the happy path where you're like, OK, we're going to get to a society that's more functioning, and then maybe paint me a darker path.

Alvin Graylin (0:29)

Yeah, I mean, nobody won the electricity race.

Nate B. Jones (0:32)

Yeah, we all got electricity.

Alvin Graylin (0:34)

Yes. But that, to me, it's scary, because if your stuff is that good, why the heck do you need, you know...

Nate B. Jones (0:41)

That much lobbying power? No, it's already here.

Alvin Graylin (0:43)

Sounds like the AI that we're creating. Essentially, we don't need to depend on the kindness of an alien species. We are creating our own alien species.

Nate B. Jones (0:52)

We are.

I've got something really special for today's video. I've been trying to find someone who can talk about US-China AI without picking a side, and I found it. Alvin Graylin has been in AI all the way back to 1990 on both sides of the Pacific. He has perspective I haven't seen anywhere else. He tells me the arms race isn't real. That's right. So stay tuned. This is going to be a fun one.

Alvin Graylin (1:25)

So I've been in tech since 1980, and I've been coding since then. I started doing AI in 1990, 1991 I was doing neural networks, controlling, creating my own red boards and doing natural language processing. And have essentially worked through the entire AI stack, everything from data centers to chip design to devices to apps, models. So I've seen a few things happen. And in fact, when I was at Intel, I worked on the chip that was the first single instruction multi-data, SIMD, which is what GPUs are based on. It was called MMX at the time. So I worked on both sides of the Pacific.

I studied at the University of Washington, and also I did my computer science at MIT and my MBA at MIT.

Recently and now, I've essentially moved from industry to more academia. I'm doing research at Stanford, at the HAI, Human Center AI Institute, and then also at, I'm teaching here at UW.

I'll be starting in the winter, but in AI and technology policy. And then I work at a think tank in DC, advising our governments on making some better policy decisions around AI and technology.

Nate B. Jones (2:40)

In that world, what do you think is the biggest misconception that people have around AI that you spend a lot of time and a lot of breath trying to defuse or change perspectives on?

Alvin Graylin (2:54)

I think the biggest problem, in fact, probably the core problem that is driving a lot of bad decisions right now is this misperception of this AI arms race.

And the expectation that there's a finish line, that there is a single winner, the winner takes all, and that the world is zero sum. And none of that is true.

Nate B. Jones (3:15)

Yeah.

Alvin Graylin (3:15)

Right. And I think when people wake up and finally understand these none of those are true, then everything that we've been doing, whether from an industry perspective or from a national security perspective, is really a misallocation of resources. The competitive nature also is preventing us from making these products and these technologies more safe.

Yeah. I mean, nobody won the electricity race. We all got electricity. And in a lot of ways, this is what, you know, you can't keep people from having electricity, and you can't keep people from having AI. It's just, it's software. And software wants to be free. Software wants to be everywhere. Software is impossible to prevent it from being distributed, especially when we start having now, you know, open source or open weight software.

Nate B. Jones (4:10)

That's right.

Alvin Graylin (4:10)

It's completely open source. It's tiny. It has a new architecture that is actually now can run on a.

Except maybe for a couple of companies who are reliant on high-profit margins from this, everybody else is good for.

Nate B. Jones (4:43)

That's right. I believe in a future where intelligence is ambient.

Alvin Graylin (4:47)

Everywhere.

Nate B. Jones (4:48)

essentially free.

Alvin Graylin (4:49)

And now it's able to run on a PC, essentially, in Mac Studio. In another few months, somebody is going to quantize it, and then it's going to run on a laptop, average laptop, and then another three, four, six months from there, it's going to run on our phones. And the level of intelligence we have in these systems is already amazing. It's more than most people can use most of the time.